Military Expenses

    

Content¶

  • Basic Conversion
  1. Overall countries
  2. Individual Country
  3. ML Models

Importing Libraries¶

In [1]:
import numpy as np # used for handling numbers
import pandas as pd # used for handling the dataset
import matplotlib.pyplot as plt
import plotly.express as px
from plotly.subplots import make_subplots
import plotly.graph_objects as go
%matplotlib inline
import seaborn as sns #data visualization library
import warnings
warnings.filterwarnings('ignore')
In [2]:
df = pd.read_csv('Military Expenditure.csv')
df.head()
Out[2]:
Name Code Type Indicator Name 1960 1961 1962 1963 1964 1965 ... 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
0 Aruba ABW Country Military expenditure (current USD) NaN NaN NaN NaN NaN NaN ... NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN
1 Afghanistan AFG Country Military expenditure (current USD) NaN NaN NaN NaN NaN NaN ... 2.518695e+08 2.981469e+08 3.258070e+08 2.385834e+08 2.171941e+08 2.682271e+08 1.995186e+08 1.858783e+08 1.914071e+08 1.980863e+08
2 Angola AGO Country Military expenditure (current USD) NaN NaN NaN NaN NaN NaN ... 3.311193e+09 3.500795e+09 3.639496e+09 4.144635e+09 6.090752e+09 6.841864e+09 3.608299e+09 2.764055e+09 3.062873e+09 1.983614e+09
3 Albania ALB Country Military expenditure (current USD) NaN NaN NaN NaN NaN NaN ... 1.827369e+08 1.858932e+08 1.970068e+08 1.832047e+08 1.800155e+08 1.781204e+08 1.323507e+08 1.308532e+08 1.443827e+08 1.804887e+08
4 Andorra AND Country Military expenditure (current USD) NaN NaN NaN NaN NaN NaN ... NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN

5 rows × 63 columns

In [3]:
df.shape
Out[3]:
(264, 63)

We check the percentage of types we have

In [4]:
df['Type'].value_counts(normalize = True) * 100
Out[4]:
Country                           76.515152
Regions Clubbed Economically      12.500000
Semi Autonomous Region             6.818182
Regions Clubbed Geographically     4.166667
Name: Type, dtype: float64

Overview on NULL values¶

In [5]:
df_null_check = pd.DataFrame(df.isnull().sum().iloc[5:]).transpose()
series_null_percent = df.isnull().sum().apply(lambda x : (x/df.shape[0])*100)
df_null_percent = pd.DataFrame(series_null_percent)
df_null_percent.rename(columns = {"":"columns",0:"null_percent"},inplace = True)
df_null_percent
Out[5]:
null_percent
Name 0.000000
Code 0.000000
Type 0.000000
Indicator Name 0.000000
1960 70.075758
... ...
2014 23.106061
2015 25.000000
2016 25.378788
2017 26.136364
2018 25.757576

63 rows × 1 columns

In [6]:
fig,ax = plt.subplots(figsize=(30,1))
ax.tick_params(labelsize=20)
sns.heatmap(df_null_percent.transpose().iloc[:,4:],cmap = 'coolwarm',cbar=False)
plt.xticks(rotation=45,)
plt.yticks(rotation=0)
plt.show()
Result:
NULL Values are far greater in earlier years. Ths might be due to lack of technological advancements.

Tasking For Major Countries In The Dataset¶

We create a new Total column as a sum of consumption over the years for each country

In [8]:
df1 = df.assign(Total = df.sum(axis=1))
df1.head(5)
Out[8]:
Name Code Type Indicator Name 1960 1961 1962 1963 1964 1965 ... 2010 2011 2012 2013 2014 2015 2016 2017 2018 Total
0 Aruba ABW Country Military expenditure (current USD) NaN NaN NaN NaN NaN NaN ... NaN NaN NaN NaN NaN NaN NaN NaN NaN 0.000000e+00
1 Afghanistan AFG Country Military expenditure (current USD) NaN NaN NaN NaN NaN NaN ... 2.981469e+08 3.258070e+08 2.385834e+08 2.171941e+08 2.682271e+08 1.995186e+08 1.858783e+08 1.914071e+08 1.980863e+08 3.239602e+09
2 Angola AGO Country Military expenditure (current USD) NaN NaN NaN NaN NaN NaN ... 3.500795e+09 3.639496e+09 4.144635e+09 6.090752e+09 6.841864e+09 3.608299e+09 2.764055e+09 3.062873e+09 1.983614e+09 7.011220e+10
3 Albania ALB Country Military expenditure (current USD) NaN NaN NaN NaN NaN NaN ... 1.858932e+08 1.970068e+08 1.832047e+08 1.800155e+08 1.781204e+08 1.323507e+08 1.308532e+08 1.443827e+08 1.804887e+08 3.054373e+09
4 Andorra AND Country Military expenditure (current USD) NaN NaN NaN NaN NaN NaN ... NaN NaN NaN NaN NaN NaN NaN NaN NaN 0.000000e+00

5 rows × 64 columns

We replace NaN values with zero

This was done to make sure that plotting had no errors
Why didn't we impute?
Imputing factual Data of militiary which it self is a sensitive topic could have lead to mis-leading results. We don't want data that will hinder our actions just because the analysis was wrong. "I rather be uninformed than to be misinformed."

In [9]:
df1.fillna(0, inplace=True)
df1.head(5)
Out[9]:
Name Code Type Indicator Name 1960 1961 1962 1963 1964 1965 ... 2010 2011 2012 2013 2014 2015 2016 2017 2018 Total
0 Aruba ABW Country Military expenditure (current USD) 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
1 Afghanistan AFG Country Military expenditure (current USD) 0.0 0.0 0.0 0.0 0.0 0.0 ... 2.981469e+08 3.258070e+08 2.385834e+08 2.171941e+08 2.682271e+08 1.995186e+08 1.858783e+08 1.914071e+08 1.980863e+08 3.239602e+09
2 Angola AGO Country Military expenditure (current USD) 0.0 0.0 0.0 0.0 0.0 0.0 ... 3.500795e+09 3.639496e+09 4.144635e+09 6.090752e+09 6.841864e+09 3.608299e+09 2.764055e+09 3.062873e+09 1.983614e+09 7.011220e+10
3 Albania ALB Country Military expenditure (current USD) 0.0 0.0 0.0 0.0 0.0 0.0 ... 1.858932e+08 1.970068e+08 1.832047e+08 1.800155e+08 1.781204e+08 1.323507e+08 1.308532e+08 1.443827e+08 1.804887e+08 3.054373e+09
4 Andorra AND Country Military expenditure (current USD) 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00

5 rows × 64 columns

We divide the values by 10 to the power 9 to get the values in billions of USD (it's more clear)

In [10]:
columns=[str(i) for i in list((range(1960,2019)))]
columns=columns+["Total"]
for i in columns:
    df1[i]=df1[i]/1.e+9
df1=np.round(df1, decimals=2)
df1.head()
Out[10]:
Name Code Type Indicator Name 1960 1961 1962 1963 1964 1965 ... 2010 2011 2012 2013 2014 2015 2016 2017 2018 Total
0 Aruba ABW Country Military expenditure (current USD) 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
1 Afghanistan AFG Country Military expenditure (current USD) 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.30 0.33 0.24 0.22 0.27 0.20 0.19 0.19 0.20 3.24
2 Angola AGO Country Military expenditure (current USD) 0.0 0.0 0.0 0.0 0.0 0.0 ... 3.50 3.64 4.14 6.09 6.84 3.61 2.76 3.06 1.98 70.11
3 Albania ALB Country Military expenditure (current USD) 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.19 0.20 0.18 0.18 0.18 0.13 0.13 0.14 0.18 3.05
4 Andorra AND Country Military expenditure (current USD) 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

5 rows × 64 columns

Sorting by total consumption for type Country

In [11]:
df1.sort_values(by=['Type','Total'],ascending=[False,False],inplace=True)
df1=df1[df1['Type'].str.contains("Country")]

Top 20 countries by total consumption with unnecessary attributes removed

In [12]:
df2 = df1[:20]
df3 = df2.drop(['Indicator Name', 'Code', 'Type'], axis=1)
new = df3.reset_index(drop=True)
new.head(20)
Out[12]:
Name 1960 1961 1962 1963 1964 1965 1966 1967 1968 ... 2010 2011 2012 2013 2014 2015 2016 2017 2018 Total
0 United States 45.38 47.81 52.38 52.30 51.21 51.83 63.57 75.45 80.73 ... 698.00 711.00 685.00 640.00 610.00 596.00 600.00 606.00 649.00 17284.12
1 China 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 ... 116.00 138.00 157.00 180.00 201.00 214.00 216.00 228.00 250.00 2368.37
2 France 3.88 4.13 4.49 4.63 4.92 5.12 5.41 5.86 6.13 ... 61.78 64.60 60.04 62.42 63.61 55.34 57.36 60.42 63.80 1927.28
3 United Kingdom 4.59 4.75 5.01 5.20 5.51 5.79 5.98 6.20 5.56 ... 58.08 60.27 58.50 56.86 59.18 53.86 48.12 46.43 50.00 1801.94
4 Germany 2.88 3.27 4.31 4.98 4.89 4.98 5.06 5.35 4.83 ... 46.26 48.14 46.47 45.93 46.10 39.81 41.58 45.38 49.47 1611.06
5 Japan 0.48 0.49 0.54 0.63 0.72 0.82 0.93 1.04 1.17 ... 54.66 60.76 60.01 49.02 46.88 42.11 46.47 45.39 46.62 1511.24
6 Saudi Arabia 0.14 0.07 0.08 0.11 0.12 0.18 0.30 0.34 0.34 ... 45.24 48.53 56.50 67.02 80.76 87.19 63.67 70.40 67.55 1241.87
7 Russian Federation 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 ... 58.72 70.24 81.47 88.35 84.70 66.42 69.25 66.53 61.39 1014.83
8 India 0.68 0.75 1.07 1.80 1.99 2.13 1.66 1.49 1.59 ... 46.09 49.63 47.22 47.40 50.91 51.30 56.64 64.56 66.51 935.60
9 Italy 1.01 1.06 1.22 1.46 1.59 1.72 1.90 1.93 1.99 ... 32.02 33.83 29.78 29.96 27.70 22.18 25.03 26.45 27.81 903.51
10 South Korea 0.28 0.16 0.19 0.19 0.14 0.13 0.18 0.22 0.28 ... 28.18 30.99 31.95 34.31 37.55 36.57 36.89 39.17 43.07 719.98
11 Brazil 0.38 0.34 0.39 0.44 0.35 0.63 0.72 0.75 0.76 ... 34.00 36.94 33.99 32.87 32.66 24.62 24.22 29.28 27.77 593.44
12 Canada 1.70 1.68 1.67 1.61 1.66 1.57 1.61 1.78 1.80 ... 19.32 21.39 20.45 18.52 17.85 17.94 17.78 21.34 21.62 522.83
13 Spain 0.23 0.24 0.31 0.32 0.39 0.44 0.56 0.55 0.67 ... 19.71 19.70 18.86 17.24 17.18 15.19 14.01 16.04 18.25 509.99
14 Australia 0.46 0.47 0.49 0.55 0.66 0.79 0.99 1.20 1.32 ... 23.22 26.60 26.22 24.83 25.78 24.05 26.38 27.69 26.71 508.91
15 Iran 0.10 0.11 0.11 0.13 0.16 0.20 0.26 0.33 0.44 ... 13.56 14.28 16.49 12.00 9.90 10.59 12.26 13.93 13.19 433.48
16 Israel 0.19 0.24 0.21 0.26 0.28 0.31 0.38 0.61 0.69 ... 14.61 16.34 15.57 17.32 18.49 16.97 14.78 15.58 15.95 430.63
17 Turkey 0.47 0.30 0.33 0.35 0.38 0.42 0.44 0.51 0.57 ... 17.94 17.30 17.96 18.66 17.77 15.88 17.85 17.82 18.97 400.51
18 Poland 3.55 4.03 4.38 4.95 5.20 5.52 5.98 6.28 7.22 ... 8.79 9.46 8.99 9.28 10.35 10.21 9.16 9.87 11.60 357.68
19 Netherlands 0.45 0.55 0.60 0.64 0.74 0.75 0.77 0.88 0.91 ... 11.22 11.65 10.36 10.23 10.33 8.67 9.12 9.58 11.24 343.92

20 rows × 61 columns

Visualizing the clean data

In [13]:
plt.figure(figsize=(12,8))
sns.barplot(x = 'Total', y = 'Name', data = df3)
plt.title('Total Millitary Spending from 1960 to 2018')
plt.xlabel('Total in bilions USD')
plt.ylabel('Countries')
plt.grid()
In [14]:
fig = px.pie(df2, values='Total', names='Name', title='Total military spendings in percentage from 1960 to 2018 ')
fig.show()
In [15]:
fig = px.pie(df2, values='1960', names='Name', title='Military spendings in percentage in 1960')
fig.show()

Plotting maps for better feel of the Data

In [16]:
fig = px.scatter_geo(df2, locations = 'Code',hover_name="Name",size = '2018')
fig.update_layout(title="First 20 most powerful country")
fig.show()
In [17]:
df4 = df3.drop(['Total'], axis=1)
In [18]:
Top20 = df4.set_index('Name')
Top20.index = Top20.index.rename('Year')
Top20 = Top20.T
Top20.head()
Out[18]:
Year United States China France United Kingdom Germany Japan Saudi Arabia Russian Federation India Italy South Korea Brazil Canada Spain Australia Iran Israel Turkey Poland Netherlands
1960 45.38 0.0 3.88 4.59 2.88 0.48 0.14 0.0 0.68 1.01 0.28 0.38 1.70 0.23 0.46 0.10 0.19 0.47 3.55 0.45
1961 47.81 0.0 4.13 4.75 3.27 0.49 0.07 0.0 0.75 1.06 0.16 0.34 1.68 0.24 0.47 0.11 0.24 0.30 4.03 0.55
1962 52.38 0.0 4.49 5.01 4.31 0.54 0.08 0.0 1.07 1.22 0.19 0.39 1.67 0.31 0.49 0.11 0.21 0.33 4.38 0.60
1963 52.30 0.0 4.63 5.20 4.98 0.63 0.11 0.0 1.80 1.46 0.19 0.44 1.61 0.32 0.55 0.13 0.26 0.35 4.95 0.64
1964 51.21 0.0 4.92 5.51 4.89 0.72 0.12 0.0 1.99 1.59 0.14 0.35 1.66 0.39 0.66 0.16 0.28 0.38 5.20 0.74

Race Graph derivation idea

In [19]:
plt.figure(figsize=(20,10))
plt.plot(Top20.index, Top20.values)
plt.ylabel('Spendings through year')
plt.title('Top 20 Countries  in Military Expenditure ')
plt.xticks(rotation=45)
plt.legend(Top20.columns)
plt.grid(True)
plt.show()
In [20]:
# Percentage of consumption growth by year for the top 20 countries

# general code format
# perccode = (Top20[countryname].iloc[-1] - Top20[countryname].iloc[0]) * 100 / Top20[countryname].iloc[0]

PercUSA = (Top20['United States'].iloc[-1] - Top20['United States'].iloc[0])*100/Top20['United States'].iloc[0]
PercChina = (Top20['China'].iloc[-1] - Top20['China'].iloc[29])*100/Top20['China'].iloc[29]
PercRUS = (Top20['Russian Federation'].iloc[-1] - Top20['Russian Federation'].iloc[33])*100/Top20['Russian Federation'].iloc[33]
PercISR = (Top20['Israel'].iloc[-1] - Top20['Israel'].iloc[0])*100/Top20['Israel'].iloc[0]
PercITA = (Top20['Italy'].iloc[-1] - Top20['Italy'].iloc[0])*100/Top20['Italy'].iloc[0]
PercJPN = (Top20['Japan'].iloc[-1] - Top20['Japan'].iloc[0])*100/Top20['Japan'].iloc[0]
PercNET = (Top20['Netherlands'].iloc[-1] - Top20['Netherlands'].iloc[0])*100/Top20['Netherlands'].iloc[0]
PercPOL = (Top20['Poland'].iloc[-1] - Top20['Poland'].iloc[0])*100/Top20['Poland'].iloc[0]
PercSAU = (Top20['Saudi Arabia'].iloc[-1] - Top20['Saudi Arabia'].iloc[0])*100/Top20['Saudi Arabia'].iloc[0]
PercKOR = (Top20['South Korea'].iloc[-1] - Top20['South Korea'].iloc[0])*100/Top20['South Korea'].iloc[0]
PercSPA = (Top20['Spain'].iloc[-1] - Top20['Spain'].iloc[0])*100/Top20['Spain'].iloc[0]
PercTUR = (Top20['Turkey'].iloc[-1] - Top20['Turkey'].iloc[0])*100/Top20['Turkey'].iloc[0]
PercUK = (Top20['United Kingdom'].iloc[-1] - Top20['United Kingdom'].iloc[0])*100/Top20['United Kingdom'].iloc[0]
PercAUS = (Top20['Australia'].iloc[-1] - Top20['Australia'].iloc[0])*100/Top20['Australia'].iloc[0]
PercBRA = (Top20['Brazil'].iloc[-1] - Top20['Brazil'].iloc[0])*100/Top20['Brazil'].iloc[0]
PercCAN = (Top20['Canada'].iloc[-1] - Top20['Canada'].iloc[0])*100/Top20['Canada'].iloc[0]
PercFRA = (Top20['France'].iloc[-1] - Top20['France'].iloc[0])*100/Top20['France'].iloc[0]
PercGER = (Top20['Germany'].iloc[-1] - Top20['Germany'].iloc[0])*100/Top20['Germany'].iloc[0]
PercIND = (Top20['India'].iloc[-1] - Top20['India'].iloc[0])*100/Top20['India'].iloc[0]
PercIRA = (Top20['Iran'].iloc[-1] - Top20['Iran'].iloc[0])*100/Top20['Iran'].iloc[0]
In [21]:
data = [['United States', PercUSA], ['China', PercChina], ['France', PercFRA], ['United Kingdom', PercUK], ['Germany', PercGER], ['Japan', PercJPN], ['Saudi Arabia', PercSAU], ['Russian Federation', PercRUS], ['India', PercIND], ['Italy', PercITA], ['South Korea', PercKOR], ['Brazil', PercBRA], ['Canada', PercCAN], ['Spain', PercSPA], ['Australia', PercAUS], ['Iran', PercIRA], ['Israel', PercISR], ['Turkey', PercTUR], ['Poland', PercPOL], ['Netherlands', PercNET]]
In [22]:
percdf= pd.DataFrame(data, columns=['Country', 'Percentage growth'])
percdf.head(20)
Out[22]:
Country Percentage growth
0 United States 1330.145439
1 China 2092.982456
2 France 1544.329897
3 United Kingdom 989.324619
4 Germany 1617.708333
5 Japan 9612.500000
6 Saudi Arabia 48150.000000
7 Russian Federation 690.090090
8 India 9680.882353
9 Italy 2653.465347
10 South Korea 15282.142857
11 Brazil 7207.894737
12 Canada 1171.764706
13 Spain 7834.782609
14 Australia 5706.521739
15 Iran 13090.000000
16 Israel 8294.736842
17 Turkey 3936.170213
18 Poland 226.760563
19 Netherlands 2397.777778
In [23]:
# Barplot of percentage growth of the top 20 countries plotly

fig = px.bar(percdf, x='Country', y='Percentage growth', title='Percentage growth of the top 20 countries')
fig.show()
In [24]:
model = percdf.join(new['Total'])
model
Out[24]:
Country Percentage growth Total
0 United States 1330.145439 17284.12
1 China 2092.982456 2368.37
2 France 1544.329897 1927.28
3 United Kingdom 989.324619 1801.94
4 Germany 1617.708333 1611.06
5 Japan 9612.500000 1511.24
6 Saudi Arabia 48150.000000 1241.87
7 Russian Federation 690.090090 1014.83
8 India 9680.882353 935.60
9 Italy 2653.465347 903.51
10 South Korea 15282.142857 719.98
11 Brazil 7207.894737 593.44
12 Canada 1171.764706 522.83
13 Spain 7834.782609 509.99
14 Australia 5706.521739 508.91
15 Iran 13090.000000 433.48
16 Israel 8294.736842 430.63
17 Turkey 3936.170213 400.51
18 Poland 226.760563 357.68
19 Netherlands 2397.777778 343.92

Plotting the correlation between the percentage growth and the total military spendings using plotly

In [25]:
fig = px.scatter(model, x="Percentage growth", y="Total", color="Country", size='Total', hover_name="Country", log_x=True, size_max=60)
fig.show()
In [26]:
fig = px.sunburst(model, path=['Country', 'Percentage growth'], values='Total', color='Percentage growth', color_continuous_scale='rdbu')
fig.update_layout(margin = dict(t=10, l=0, r=0, b=0))
fig.show()

Individual Countries¶

Tool for Indiviual Countries

    

Analyzing a countries military spending with historical context.


For the time being we are considering the country as India

Let's select only data only related to the country and also drop some columns that will have no effect in our analysis, such as "Code", "Type" and "Indicator Name" We will take example of India for our country for now.

In [27]:
df_ind = df[df.Name == "India"]
df_ind = df_ind.drop(["Code", "Type", "Indicator Name"], axis=1)
df_ind = df_ind.set_index("Name")
df_ind
Out[27]:
1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 ... 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
Name
India 681765681.8 748388248.4 1.065436e+09 1.795449e+09 1.986654e+09 2.125990e+09 1.661155e+09 1.487733e+09 1.586867e+09 1.691433e+09 ... 3.872215e+10 4.609045e+10 4.963382e+10 4.721692e+10 4.740353e+10 5.091411e+10 5.129548e+10 5.663762e+10 6.455944e+10 6.651029e+10

1 rows × 59 columns

Checking for null values:

In [28]:
df_ind.isnull().sum().sum()
Out[28]:
0

Looks good. We can also transpose the dataset to make it easier to work with.

In [29]:
df_ind = df_ind.transpose()

Now we can plot this data and visualize the growth of India's military spending over these years:

In [30]:
df_ind = df_ind.div(10**9) #making biger numbers smaller.
fig = plt.figure()
ax = fig.add_axes([0,0,2.2,1.5])
plt.plot(df_ind['India'],color = '#0B5345');
font1 = {'family':'times new roman','color':'blue','size':40}
font2 = {'family':'serif','color':'#D35400','size':25}
plt.title("India's military Budget in Billion USD$",fontdict = font1)
plt.xlabel('Years',fontdict = font2)
plt.ylabel('Expenditure', fontdict = font2)
plt.grid()
plt.xticks(rotation = 45)
plt.show()

Our data ranges from 1960 to 2018. We can start drilling down by singleling out each decade

In [31]:
decades = ["196", "197", "198", "199", "200", "201"]
rows = 2
cols = 3
fig, axes = plt.subplots(nrows=rows, ncols=cols, figsize=(20,11), sharey=True)
fig.suptitle("Indian military budget by year in billions of US$", fontsize=40, color = '#800000')

for i, decade in enumerate(decades):
    a, b = divmod(i, cols)
    df_decade = df_ind[df_ind.transpose().columns.str.startswith(decade)]
    df_decade.div(10**9).plot.bar(ax=axes[a,b], legend=False, rot=45,color = '#5B2C6F')
plt.show()

It is really well worth citing that the grouped bar graphs above percentage a not unusual place y scale. This makes it smooth to visualize the boom of the records thru the to be had period, like we did within side the line graph above, however makes it difficult to visualize smaller intervals of boom. This is in particular due to the massive spike in our navy price range within side the early 2000s. We'll get to that point, however first, let's cut up our former graphs in groups: pre-2000s and post-2000s, and notice if any new developments appear

In [32]:
_df = df_ind.transpose()
df_pre_2k = _df.loc[:, _df.columns.str.startswith("19")].transpose()
df_pos_2k = _df.loc[:, _df.columns.str.startswith("20")].transpose()

Pre_2000s¶

In [33]:
df_pre_2k.div(10**9).plot(figsize=(12,8), legend=False, rot=45, color = '#5B2C6F');
In [34]:
_years = tuple([str(y) for y in list(range(1984, 2000))])
df_80s = df_pre_2k.transpose()
df_80s = df_80s.loc[:,'1984':'2000'].transpose().div(10)
df_80s.div(10**9).plot.bar(figsize=(10,8), legend=False, rot=0, title="Military spending in India in billions of US$");
In [35]:
pct_serie = df_80s.transpose().iloc[0].pct_change()
pct_serie[1:].multiply(100).plot.bar(figsize=(10,8), rot=0, color=(pct_serie[1:] > 0).map({True: 'g', False: 'C3'}));
plt.title("Percentage change in military spending in India from 2001 to 2018", fontsize=15, color = '#800000', fontweight='bold')
plt.xlabel('Years', fontsize=20 )
plt.ylabel('Growth', fontsize=20)
plt.show()
In [36]:
budgets_comp = df_80s.transpose()[["1984", "1987"]].iloc[0].pct_change()[1]
budgets_comp
Out[36]:
0.5635910139269893

Status of buget before 2000s¶

We as Indians took our freedom in 1947, During the first 2 decades we were heavily recovering from the economical and structure damages.. But even then we started focusing on the military aspect of a country.

-> 24 May 1985 - Terrorist and Disruptive Activities (Prevention) Act is came into effect after presidential assent.

-> 30 April 1986 – Operation Black Thunder conducted to flush out remaining Sikh extremists from the Golden Temple.

-> 1991 Economic Crisis

-> 1996 A devastating Category 4 Cyclone strikes Andhra Pradesh, India. The storm surge sweeps fishing villages out to sea, over 2,000 people die. 95% of the crops are completely destroyed.

-> 1998 :- First Atomic test of India (Pokhran,rajasthan), Over 1,120 people are killed when a cyclone hits coastal areas of the western state of Gujarat.

2000s¶

In [37]:
df_pos_2k.div(10**9).plot(figsize=(12,8), legend=False, color = '#800000');
# adding title and axis names
plt.title("India's military budget in billions of US$",fontsize=25, color = '#800000', fontweight='bold')
plt.xlabel('Years', fontsize=20 )
plt.ylabel('Expenditure', fontsize=20)
plt.show()
In [38]:
_years = tuple([str(y) for y in list(range(2000, 2019))])
df_2ks = df_pos_2k.transpose()
df_2ks = df_2ks.loc[:, df_2ks.columns.str.startswith(_years)].transpose().div(10**9)
df_2ks.plot.bar(figsize=(10,8), legend=False,rot = 0,);
plt.title("Military spending in billions of US$",fontsize=25, color = '#800000', fontweight='bold')
plt.xlabel('Years', fontsize=20 )
plt.ylabel('Expenditure', fontsize=20)
plt.show()

Change in percentage after year 2000

In [39]:
pct_serie = df_2ks.transpose().iloc[0].pct_change()
pct_serie[1:].multiply(100).plot.bar(figsize=(10,8), rot=0,  color=(pct_serie[1:] > 0).map({True: 'g', False: 'r'}));
plt.title("Percentage change in military spending in India from 2001 to 2018", fontsize=15, color = '#800000', fontweight='bold')
plt.xlabel('Years', fontsize=20 )
plt.ylabel('Growth', fontsize=20)
plt.show()

Status of buget after 2000s¶

-> 2002 - 2003 Finance Minister Yashwant Sinha presents the 2002–03 budget. Amongst its major features are a 4.8% increase in defence spending and a 5% surcharge on income tax to pay for this.

-> 2005 Mumbai and the Mumbai Conurbation area is submerged in 5–7 ft. of water due to heavy rains, making nearby dams release water causing a massive flood, which virtually stops the financial capital of India for 4–5 days.

-> 2006 A terrorist attack in Varanasi killed at least 15 people and left more than 50 injured. A series of coordinated bomb attacks struck several commuter trains in Mumbai, India during the evening rush hour.

-> 2012 Major fire in the signalling system at Kurla brings Central line and Harbour line of Mumbai Suburban Railway to a standstill and two commuters lose their lives after falling off an overcrowded local train. The loss was expected to be ₹170 million (US$2.1 million) to the Central Railway. A pair of great earthquakes occur in the Wharton Basin west of Sumatra in Indonesia. The maximum Mercalli intensity of this strike-slip doublet earthquake was VII (Very strong). Ten were killed, twelve were injured, and a non-destructive tsunami was observed on the island of Nias.

-> 2015 ISRO launched 5 UK satellites with a total payload of 1,440 kg using the PSLV-C28 launch vehicle.

-> During the rest year the growth was pretty normal.

How this correlates with India's general budget?¶

Now to put some perspective on the visualized data:

Our last availabe data, for the year of 2018, shows that the military budget for that year was around 66.5 billions of US dollars.

In [40]:
df_ind.transpose()["2018"].India
Out[40]:
66.510289108

Indian's yearly budget is defined by it's Central Government.

In December 2017, the US dollar to Indian rupees conversion rate was around 1 / 65.0966 meaning 1 US dollar was worth around 65.0966 Indian Rupee. This means that the military budget for that year was higher than 4329 billions of indian rupees.

In [41]:
military_2018 = df_ind.transpose()["2018"].multiply(65.0966).India
military_2018
Out[41]:
4329.5936859478325

In 2018 India's budget for health and education was approximately 5666.44 and 8459.26 billions of rupees respectivly. This means that India's government is currently promoting economic status for its citizens.

In [42]:
health_2018 = 5666.44  # sourced from internet
educat_2018 = 8459.26 #sourced from internet
series_2018 = pd.Series({"Health": health_2018,"Education": educat_2018,"Military": military_2018})

fig, axes = plt.subplots(ncols=2, figsize=(18,6))
fig.suptitle("Indian military x education x health budget - 2018", fontsize=25, color = '#800000')
series_2018.sort_values(ascending=False).plot.bar(ax=axes[0], rot=0, color = ['#0B5345','#800000','#5B2C6F'])
series_2018.plot.pie(ax=axes[1], autopct="%1.2f%%", colors = ['#800000','#006400','#5B2C6F'], fontsize=15, wedgeprops={"linewidth":2,"edgecolor":"white"}, textprops={"fontsize":15},shadow=True, startangle=90);
axes[1].set_ylabel("");

In genral india is currently givin more share towards education.. Looking at its current situation, we can definitly say that it is the right step as we have no indication regarding any stress with any other country which might lead to war.

Internet resources¶

https://www.google.com/search?q=What+is+the+defence+budget+of+India+in+2018%3F&sa=X&ved=2ahUKEwifloqbwYj7AhXNQ94KHbejCQUQzmd6BAglEAU&biw=856&bih=884&dpr=1.5

https://www.google.com/search?q=india%27s+total+healthcare+budget+2018&spell=1&sa=X&ved=2ahUKEwidgtGlwIj7AhUH0GEKHUTJDZIQBSgAegQIBxAB&biw=856&bih=884&dpr=1.5

ML Models¶

In [43]:
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LinearRegression,SGDRegressor
from sklearn.preprocessing import PolynomialFeatures,StandardScaler
from sklearn.metrics import r2_score
from sklearn.pipeline import Pipeline

Convertin df into ML model useable form¶

In [44]:
df_ml = df.dropna()
df_ml
df_ml = df_ml.drop(columns= ['Code','Type','Indicator Name'])

We only took the full data from countries which had no null values. As imputation will hinder the model accuracy.¶

In [45]:
df_ml.iloc[:,1:] = df_ml.iloc[:,1:].div(10**9) #made the vales shorter for better clarity
In [46]:
df_ml.rename(columns = {'Name':'Srno'}, inplace = True)
df_ml = df_ml.set_index('Srno').transpose()
df_ml.index.astype('int64')
df_ml.reset_index(inplace=True)
df_ml.rename(columns = {'index':'Year'}, inplace = True)
In [47]:
df_ml.astype({'Year':'int64'})
df_ml['Year']= [1960, 1961, 1962, 1963, 1964, 1965, 1966, 1967, 1968,
       1969, 1970, 1971, 1972, 1973, 1974, 1975, 1976, 1977,
       1978, 1979, 1980, 1981, 1982, 1983, 1984, 1985, 1986,
       1987, 1988, 1989, 1990, 1991, 1992, 1993, 1994, 1995,
       1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004,
       2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013,
       2014, 2015, 2016, 2017, 2018]
df_ml.dtypes
Out[47]:
Srno
Year                         int64
Australia                  float64
Austria                    float64
Belgium                    float64
Burkina Faso               float64
                            ...   
South Asia (IDA & IBRD)    float64
Tunisia                    float64
Turkey                     float64
United States              float64
South Africa               float64
Length: 67, dtype: object

The data set is now cleaned and rearranged for ML model.

In [48]:
fig = plt.figure(figsize = (20,5))
plt.scatter(df_ml.index,df_ml['India'])
plt.xticks(rotation = 45)
plt.title('Expense through the year', fontsize = 30,color = 'blue',fontweight = 'bold')
plt.xlabel('Year',fontsize = 15, color = '#800080')
plt.ylabel('Expenses in billion', fontsize = 15, color = '#800080')
plt.show()

SGD Regression¶

In [49]:
# SGD regression machine learning model
X = df_ml['Year'].values.reshape(-1,1)
y = df_ml['India'].values.reshape(-1,1)
X_train,X_test,y_train,y_test = train_test_split(X,y,test_size = 0.2,random_state = 0)
scaler = StandardScaler()
X_train = scaler.fit_transform(X_train)
X_test = scaler.transform(X_test)
sgd_reg = SGDRegressor(max_iter = 1000,tol = 1e-3,penalty = None,eta0 = 0.1)
sgd_reg.fit(X_train,y_train.ravel())
y_pred = sgd_reg.predict(X_test)

# accuracy of the model
r2_score(y_test,y_pred)
Out[49]:
-1.0831786912998873
In [50]:
# plotting the SGD regression
plt.figure(figsize=(20,5))
plt.plot(X_test,y_pred,color = 'red')
plt.scatter(X_test,y_test,color = 'blue')
plt.title('SGD regression',fontsize = 30, color = 'blue',fontweight = 'bold')
plt.xlabel('Year',fontsize = 15, color = '#800080')
plt.ylabel('Expenses in billion',fontsize = 15, color = '#800080')
plt.show()

Polynomial Regression¶

In [51]:
# polynomial regression machine learning model
X = df_ml['Year'].values.reshape(-1,1)
y = df_ml['India'].values.reshape(-1,1)

# dedining train and test data points
X_train,X_test,y_train,y_test = train_test_split(X,y,test_size = 0.2,random_state = 0)
poly_reg = PolynomialFeatures(degree = 4)

# define lin_reg
lin_reg = LinearRegression()

# define pipeline
pipeline = Pipeline([('poly',poly_reg),('lin_reg',lin_reg)])
pipeline.fit(X_train,y_train)
y_pred = pipeline.predict(X_test)
X_poly = poly_reg.fit_transform(X_train)

# accuracy of the model in percentage

print("accuracy =",r2_score(y_test,y_pred)*100)

accuracy = 80.72113788055056
In [52]:
# plotting the polynomial regression model

plt.figure(figsize=(20,5))
plt.scatter(X,y,color = 'red')
plt.plot(X,lin_reg.predict(poly_reg.fit_transform(X)),color = 'blue')
plt.title('Polynomial Regression', fontsize = 30,color = 'blue',fontweight = 'bold')
plt.xlabel('Year' , fontsize = 15, color = '#800080')
plt.ylabel('Expenses in billion', fontsize = 15, color = '#800080')
plt.show()
In [53]:
# predicting values for the year 2019 and 2020

print(lin_reg.predict(poly_reg.fit_transform([[2019]])))
print(lin_reg.predict(poly_reg.fit_transform([[2020]])))

[[72.4069378]]
[[77.22599783]]
In [54]:
#plotting values with the year 2019 and 2020

plt.figure(figsize=(20,5))
plt.scatter(X,y,color = 'red')
plt.plot(X,lin_reg.predict(poly_reg.fit_transform(X)),color = 'blue')
plt.scatter(2019,lin_reg.predict(poly_reg.fit_transform([[2019]])),color = 'green')
plt.scatter(2020,lin_reg.predict(poly_reg.fit_transform([[2020]])),color = 'green')
plt.title('Polynomial Regression', fontsize = 30,color = 'blue',fontweight = 'bold')
plt.xlabel('Year' , fontsize = 15, color = '#800080')
plt.ylabel('Expenses in billion', fontsize = 15, color = '#800080')
plt.show()

Conclution

We analysed various aspects through our data related to costing and growth around the years. Finding poteintial reasons to why certain data point is acting in some way is crucial. We created a model which can somewhat predict the budget of next years. Through the process we learned a lot which can definilty help us contribute to the society.